@User uploaded a file:

Run JUnit tests

That's incorrect

Hm. Perhaps. So what’s correct?

just measure time in code?

You have to warm-up JVM

Use JMH instead: http://openjdk.java.net/projects/code-tools/jmh/

Also, you have to run your code repeatedly in a loop to measure average performance

@alexey.andreev I run code many times on different data for each measuring.

Because there are too much things that impact performance

How many times?

@alexey.andreev so I think JVM can’t be ‘warmed-up’ in this context. Every test function has its own random data.

10000

10 000 random string (consisting of A/C/T/G symbols) with length = random(9900 ~ 10000)

Anyway, measuring performance via JUnit is a bad idea, very bad

But, anyway, does somebody have an idea why Kotlin-version of code is faster in JUnit-tests when throws RuntimeException?

I mean ok, we can agree that benchmarking on JUnit - is kind of bad idea if you want to get correct-measured time of function execution. But, trend is visible - when you have “throw …” in code - it runs faster.

@egslava Could you check this sample:

val String.reverseComplement11: String get(){
    val thisLength = length - 1
    return String( CharArray(thisLength + 1) { val complementedChar = this[thisLength - it].complement1; complementedChar})
}

If you have a coffee cup with sugar in Porsche, it runs faster. Faster than tractor.

@orangy and when you doesn’t have - will it run slower? 🙂

Seriously, don’t measure things like this with JUnit timing and claim any trends. Do proper testing with JMH. Even with that, you will need to investigate and interpret results.

btw, it seems like it’s JUnit-benchmarking fail 🙂

@orangy okay, I’ll take a look. But it was soooo comfatable to measure with JUnit 😞 You just write test and see performance.

It’s not benchmarking, it’s timing.

You see some arbitrary time. Not performance 🙂

What’s difference between time and performance?

Setting up JMH is quite easy, just use mvn archetype.

(I run this code 10000 times and use different data so JVM is not cold)

Well, that’s more a word playing, but time you see in IDEA is time between signals it receives via some kind of RPC from the other process (junit test process). So many things can come to play here, it can’t really be considered measurement. Just a very rough idea about how long the test runs. Very rough.

@User uploaded a file: and commented: @orangy @egslava Sometimes im one my project we measured performance